Jackrong

/

Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-GGUF

like 600

Please upgrade your Ollama to the latest version.

Hmm... Actually, my Ollama is the latest version 0.17.4,

Maybe Ollama has some conflicts with its underlying components. You can try running inference with LM Studio instead — the community has tested both Q4 and Q8 quantizations and confirmed they work fine.

OK，I'll have a try~

I also get:
"500 Internal Server Error: unable to load model: C:\Users*****.ollama\models\blobs\sha256-884bc358ea21ccba6e26761675c20ac0d06653e89e6c2bc211ba101d3c6977b7"

Thanks for reporting this.
I’ll re-fine-tune and repackage the model today. I originally completed the fine-tuning the day after the base model was released, so there may be some conflicts with underlying libraries or dependencies that have since changed. I’ll rebuild the package using the latest versions to improve compatibility. I’ve also noticed that some Ollama users have encountered a few bugs recently, so I’m looking into that as well. Apologies for the inconvenience, and thanks for your patience.

Thanks for reporting this.
I’ll re-fine-tune and repackage the model today. I originally completed the fine-tuning the day after the base model was released, so there may be some conflicts with underlying libraries or dependencies that have since changed. I’ll rebuild the package using the latest versions to improve compatibility. I’ve also noticed that some Ollama users have encountered a few bugs recently, so I’m looking into that as well. Apologies for the inconvenience, and thanks for your patience.

Could you please share the finetune code for learning, thank you very much

Hello, when I use the latest version of Mac Ollama to call it, there is no chat option, which makes it impossible for me to use openclaw. Is this a configuration problem, or are there any precautions I should take?

Hi, is anyone else having this issue? When I run it, I get a 400 error. I'm using Ollama ROCm.

Ollama only supports text mode for qwen35/qwe35moe for now for third party GGUFs (their own quants, from their official repo, work fine tho), until a PR that will update llama.cpp backend is merged. So for now, you need to create a Modelfile without mmproj and create the model, something like this (tested with another GGUF from HF):

# point to downloaded GGUF
FROM ./Qwen3.5-27B.Q4_K_M.gguf

# use Ollama engine
TEMPLATE {{ .Prompt }}
RENDERER qwen3.5
PARSER qwen3.5

# suggested parameters for the official model, you may tweak it
PARAMETER top_p 0.95
PARAMETER presence_penalty 1.5
PARAMETER temperature 1
PARAMETER top_k 20

LICENSE "                                Apache License
                           Version 2.0, January 2004
                        http://www.apache.org/licenses/
"

If you use the HF link, like ollama pull hf.co/Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-GGUF:Q4_K_M it may include the MMPROJ file and it will fail to start. Server will report that qwen35 architecture is not supported, although it works fine without vision and with their own model, of course. Make sure you have the latest Ollama version too.

Hi, is anyone else having this issue? When I run it, I get a 400 error. I'm using Ollama ROCm.

I can reproduce this error. Same for pullhf.co/Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-GGUF:Q8_0

I have the same error:

ollama run hf.co/Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-GGUF:Q8_0
pulling manifest
pulling 283f5c4dac2b: 100% ▕██████████████████████████████████████████████████████████████████████████▏ 28 GB
Error: 400:

Hi, is anyone else having this issue? When I run it, I get a 400 error. I'm using Ollama ROCm.

I can reproduce this error. Same for pullhf.co/Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-GGUF:Q8_0

I have the same error:

ollama run hf.co/Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-GGUF:Q8_0
pulling manifest
pulling 283f5c4dac2b: 100% ▕██████████████████████████████████████████████████████████████████████████▏ 28 GB
Error: 400:

Pull the GGUF directly from HF and then create a Modelfile to use the model.